WO2019235131A1

WO2019235131A1 - Search text generation system and search text generation method

Info

Publication number: WO2019235131A1
Application number: PCT/JP2019/018693
Authority: WO
Inventors: 拓生逸見; 佐藤　淳一
Original assignee: 株式会社ユニバーサルエンターテインメント
Priority date: 2018-06-04
Filing date: 2019-05-10
Publication date: 2019-12-12
Also published as: US20210263915A1; CN112236768A; PH12020552089A1; JP7139157B2; JP2019211987A

Abstract

A problem addressed by the present invention is to provide a search text generation system capable of automatically generating an appropriate search text from natural language input by a user. The search text generation server 300: analyzes received input text; identifies a word corresponding to an operator and a word of interest which is adjacent to said word; and, using the words having thus been identified, automatically generates a search text which corresponds to the input text.

Description

Search sentence generation system and search sentence generation method

The present invention relates to a search sentence generation system that generates a search sentence based on a natural sentence input by a user.

In recent years, a lot of information has been released on the Internet, and an Internet user can input a keyword into a search engine and specify a desired WEB page from among a huge number of WEB pages on the Internet. You can browse the necessary information.

Also, a user of a commercial database or the like can create a search expression including keywords, logical operators, parentheses, etc., and obtain target information from the commercial database.

However, when obtaining desired information by searching, it is complicated to input a keyword as described above and to create a search expression including keywords, logical operators, parentheses, etc. Unless you are used to it, it is difficult to obtain search results as intended.

Therefore, for example, an information retrieval apparatus that automatically creates a retrieval formula based on a natural sentence input by a user has been proposed (see Patent Document 1).

This information search device creates a search formula based on a natural text search question entered by the user. More specifically, create a search expression using operators by referring to a search term dictionary that stores synonyms, related terms, and hierarchical relationships of terms included in the search question. Use to search the database to be searched. In addition, after obtaining the search results, the user is inquired whether the number of answers is appropriate, and the search expression is recreated using the search term dictionary according to the answer result of the inquiry. The

Japanese Patent Laid-Open No. 5-128158

In the information search apparatus as described above, a search expression is created based on a search sentence composed of natural sentences. However, keywords and operators included in the search expression are appropriately changed according to the answer result of the inquiry by the user. It is configured to be.

For example, when the user determines that the number of answers of the search result is small, the information search device refers to the search term dictionary (in parallel) and adds or replaces keywords to replace the operator. Is changed from “AND” to “OR” to re-create a search expression that can obtain a large number of responses.

Such an information search apparatus is effective in assisting a user who is unfamiliar with the creation of a search sentence, but on the other hand, the true intention or nuance of the search included in the natural sentence may not be expressed correctly.

In addition, in the currently popular Internet search engines, for example, when a user inputs a search word such as “not gorgeous” or “not colorful”, “splendid” or “colored” respectively. A search using a keyword is performed, and the negative meaning of a search word input by a user may be eliminated.

Even in such an Internet search engine, as described above, the true intention and nuance of the search included in the natural sentence may not be expressed correctly.

Therefore, an object of the present invention is to provide a search sentence generation system capable of automatically generating an appropriate search sentence based on a natural sentence input by a user.

Another object of the present invention is to provide a search sentence generation system capable of generating a search sentence that reflects the true intention and nuance of the search included in the natural sentence based on the natural sentence input by the user. There is.

The present invention provides the following search sentence generation system and search sentence generation method.

The invention according to the first embodiment of the present invention has the following configuration.
An input sentence receiving means for receiving an input sentence of a natural sentence input from a user;
An input sentence analyzing means for analyzing the input sentence and dividing it into a plurality of words;
Based on the plurality of words, a word (for example, “de” or “other than”) corresponding to an operator (for example, AND or NOT) is determined, and a target word related to the determined word is determined. A determination means;
A search sentence generation system comprising: an operator corresponding to the determined word; and a search sentence generation means for generating a search sentence corresponding to the input sentence using a set of target words determined to be related to the word (For example, search sentence generation system 11).

Such a configuration of the present invention makes it possible to automatically generate an appropriate search sentence based on a natural sentence input by the user, and the user is not accustomed to creating a search sentence, Even if there is no knowledge, an intended search result can be obtained without creating a complicated search sentence.

The invention according to the second embodiment of the present invention has the following configuration in the first embodiment.
The determination means determines a word corresponding to an operator of a negative logical operation based on a predetermined determination criterion (for example, a word such as “other than” or “not” is extracted in an input sentence after analysis) In this case, an operator “NOT” is associated with the word).

With such a configuration of the present invention, an intended search result can be obtained even when a user's input sentence includes a negative condition.

The invention according to the third embodiment of the present invention has the following configuration in the first embodiment.
The search sentence generation means uses a predetermined operator based on a relationship between the operator corresponding to the determined word and the operator and the target word in each set of target words determined to be related to the word. The search sentence is generated so that the priority of the logical operation is determined (for example, in FIG. 8B, “clothes NOT blue” of the search sentence 364 and “3” of the partial search sentence 363 (3) “ When “clothes AND” is generated, the target word “clothes” of both is common, and in the search sentence 364, “clothes” is in a relationship modified with the operator “NOT” and the target word “blue”. Therefore, both are collected by “clothes”, and “(clothes NOT blue) AND” of the search sentence 365 is generated).

This configuration of the present invention makes it possible to automatically generate an appropriate search sentence based on a natural sentence input by a user.

The invention according to the fourth embodiment of the present invention has the following configuration in the first embodiment.
The determination means sequentially determines a word corresponding to an operator for the plurality of words (for example, in step S54 in FIG. 7, sequentially acquires words corresponding to the operator from the analysis result of the input sentence)
The search sentence generation unit is configured to generate a search sentence relating to the determined word each time a word corresponding to the operator is determined, and to integrate the search sentence with the generated search sentence. (For example, step S56 in FIG. 7).

The invention according to the fifth embodiment of the present invention has the following configuration.
A search statement generation method that is executed in a search statement generation server that generates a search statement based on an input sentence of a natural sentence input by a user,
An input sentence receiving step of receiving the input sentence;
An input sentence analysis step of analyzing the input sentence and dividing it into a plurality of words;
A determination step of determining a word corresponding to an operator based on the plurality of words and determining a target word related to the determined word;
A search sentence generation step of generating a search sentence corresponding to the input sentence by using an operator corresponding to the determined word and a set of target words determined to be related to the word. Search sentence generation method.

The search sentence generation system according to the present invention makes it possible to automatically generate an appropriate search sentence based on a natural sentence input by the user, and the user can create a normal search sentence without creating a complicated search sentence. A desired search result can be obtained by input that asks in conversation.

Further, the search sentence generation system according to the present invention makes it possible to generate a search sentence that reflects the true intention or nuance of the search included in the natural sentence based on the natural sentence input by the user. Even if the user is unfamiliar with the creation of the search sentence or does not have knowledge of the search formula, the user can obtain the intended search result without creating a complicated search sentence.

It is a figure which shows the outline | summary of the information provision system containing the search sentence production | generation system which concerns on 1st Embodiment of this invention. It is the flowchart which illustrated the response sentence output process in the information provision system of FIG. It is the flowchart which illustrated search text generation processing of the search text generation system concerning a 1st embodiment of the present invention. It is the figure which illustrated the graphical user interface of the conversation screen in a user terminal in the information provision system of FIG. It is the figure which illustrated the outline | summary of the search text production | generation process of the search text production | generation server which concerns on 1st Embodiment of this invention. It is the figure which showed the example of search text generation by the search text generation server which concerns on 1st Embodiment of this invention. It is the flowchart which illustrated search text generation processing of the search text generation server concerning a 2nd embodiment of the present invention. It is the figure which illustrated the outline | summary of the search text production | generation process of the search text production | generation server which concerns on 2nd Embodiment of this invention. It is a figure which shows the example of the hardware constitutions of the computer which comprises the user terminal relevant to the information provision system of this invention. It is a figure which shows the example of the hardware constitutions of the computer which comprises the search sentence production | generation server contained in the search sentence production | generation system of this invention.

First, the information providing system 1 including the search statement generation system 11 according to the first embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a diagram showing an overview of an information providing system 1 that realizes a conversation with a user by providing a response corresponding to the user's input. The information providing system 1 is configured to include a search sentence generation system 11 according to the first embodiment of the present invention.

The user 10 inputs a character string (for example, using a keyboard or a touch panel of the user terminal 100) in an input area of a predetermined WEB page displayed by the WEB browser executed on the user terminal 100. Here, the character string input by the user 10 is generally a search sentence for performing an information search on the search target, and is a natural sentence expressed in a natural language.

Here, it is assumed that a natural sentence related to such information retrieval is input from the user 10, and hereinafter, the input from the user is expressed as an “input sentence”. However, this input sentence may not be in the form of a sentence as long as it can be used for the search. Nor does it need to be a question or question, such as “What is a non-bittering drug?” For example, various forms of input sentences, such as an input sentence that forms part of a question sentence, such as “non-bitter medicine” can be accepted.

The user 10 interacts with the information providing system 1 in a conversation format, and the search result of the search executed based on the input sentence is displayed on the user terminal 100 as a response to the input sentence input by the user 10.

In the present embodiment, the WEB browser executed on the user terminal 100 is configured to display an input area for inputting an input sentence and a search result, but other applications executed on the user terminal 100 may be displayed. You may make it display said input area and a search result.

Next, the input sentence input from the user terminal 100 by the user 10 is provided to the information providing system 1 via the network 500 by the WEB browser of the user terminal 100. Here, the network 500 is a network including the Internet, for example.

The information providing system 1 is configured to include an information providing server 200 and a search statement generation system 11. The search sentence generation system 11 includes a search sentence generation server 300, and the search sentence generation server 300 includes analysis data 340 for analyzing an input sentence.

When the information providing server 200 of the information providing system 1 receives the input sentence from the user terminal 100, the information providing server 200 transmits the input sentence to the search sentence generation system 11, and receives the search sentence generated there from the search sentence generation system 11. In the search sentence generation system 11, when an input sentence is received, the search sentence generation server 300 refers to the analysis data 340 and the like and generates a search sentence based on the received input sentence.

When the information providing server 200 of the information providing system 1 receives the generated search sentence from the search sentence generating system 11, the information providing server 200 executes a search for the search target data 400 using the search sentence, and displays the search result as the user terminal 100. Send to.

In this embodiment, the search target data 400 is a database or the like arranged outside the information providing system 1, but may be data included in the information providing system 1. The search target data 400 may be a database that stores index information related to resources accessible via the Internet, for example. By searching such a database, it is possible to display on the user terminal 100 a list of WEB pages that match the search statement conditions, link information of WEB pages, contents of individual WEB pages, and the like.

Also, the search target data 400 can be configured to include products provided by the company, inventory information of the products, and the like. Thus, the search target data 400 can be configured with various contents.

Here, when the information providing system 1 is provided as usable for an unspecified user, the user 10 may be a general user who browses a WEB page, or a company or organization that provides a WEB site (products or services of For example, a general user who asks questions etc. is assumed, and the user terminal 100 is normally assumed to be a smartphone or a PC (personal computer) owned by the user 10 that can be connected to the Internet or the like.

In FIG. 1, the user 10 inputs an input sentence (for example, using the keyboard or touch panel of the user terminal 100) into the input area of the WEB site and transmits the input sentence to the information providing system 1. The system 1 may receive this input sentence by any method / route. For example, when the user 10 issues a query related to search by voice, the voice is subjected to voice recognition processing and converted into a character string, and the converted character string is provided to the information providing system 1 as an input sentence. can do. Further, it is also conceivable to convert the questionnaire related to the printed search into a character string by OCR (Optical Character Recognition), and provide the converted character string to the information providing system 1 as an input sentence.

In addition, the response sentence from the information providing system 1 can be provided to the user terminal 100 not only by the character information indicating the search result but also by various methods / routes such as graphics, voice, and video.

The information providing server 200 according to the present embodiment is a WEB server that operates to perform data transmission / reception with a WEB browser executed on the user terminal 100 using a protocol such as http, and to display a predetermined WEB page on the WEB browser. It also has a function.

Also, the information providing server 200 and the search statement generation server 300 are each shown here as one computer, but the same function can be distributed and executed by a plurality of computers. Further, the information providing server 200 and the search statement generation server 300 may be integrated and realized by one computer.

FIG. 1 shows an example in which the search sentence generation system 11 according to the first embodiment of the present invention is incorporated in the information providing system 1 as described above. However, the search sentence generation system 11 is various other systems. It is possible to use it. The search sentence generation system 11 is a system that receives an input sentence related to a search, at least a part of which is a natural sentence, and generates an appropriate search sentence based on the input sentence.

FIG. 2 is a flowchart showing the flow of response text output processing in the information providing system 1 according to the present embodiment. As described above, this process controls the information providing server 200 to search the content of the search target data 400 based on the input sentence input by the user 10 and to display the search result on the user terminal 100.

First, in step S <b> 11, the information providing server 200 determines whether or not the input sentence of the user 10 is received from the user terminal 100. If the input sentence has not been received (NO in step S11), the determination process is repeated. If the input sentence has been received (YES in step S11), the process proceeds to step S12.

In step S12, the information providing server 200 transmits the input sentence of the user 10 received from the user terminal 100 to the search sentence generation server 300, where a search sentence based on the input sentence is generated.

Next, in step S13, the generated search sentence is received from the search sentence generation server 300, and in step S14, the format of the received search sentence is adjusted according to the data management system of the search target data 400. For example, the search statement generation server 300 generates a search statement in a predetermined format, but the data management system of the search target data 400 has a specification that accepts a search statement in a format different from the format of the search statement generation server 300. If so, the format of the generated search statement is changed accordingly. For example, in a commercial database, there are cases where it is necessary to construct a search sentence in an SQL format including a SELECT sentence, and there are cases where it is necessary to construct it as a search sentence for use by an Internet search engine.

Next, in step S15, the content of the search target data 400 is searched using the adjusted search text, and the search result is acquired. The search result is, for example, a list of items that satisfy the search statement condition and link information to data representing details of each item. When the search target data 400 is a database that stores an index related to a resource that can be accessed via the Internet, the search result includes, for example, a list of WEB pages that satisfy the search statement condition and links to each WEB page. Information.

Next, in step S16, the information providing server 200 edits the received search result. For example, when the user terminal 100 transmits the input sentence of the user 10 by the WEB browser, the WEB page configured to display the contents of the search result is edited (in the html language or the like), and then in step S17, Control is performed to output the edited search result to the user terminal 100.

FIG. 3 is a flowchart showing the flow of search statement generation processing executed by the search statement generation server 300 according to the first embodiment. In this process, as described above, when an input sentence of the user 10 is received from the information providing server 200, a search sentence is generated according to a predetermined rule based on the input sentence, and the generated search sentence is provided as information. Transmit to server 200.

First, in step S31, the search statement generation server 300 determines whether or not the input statement of the user 10 has been received from the information providing server 200. If the input sentence has not been received (NO in step S31), the determination process is repeated. If the input sentence has been received (YES in step S31), the process proceeds to step S32.

In step S32, the search sentence generation server 300 analyzes the received input sentence and extracts a plurality of words from the input sentence. In the present embodiment, morpheme analysis is performed using a morpheme database included in the analysis data 340. As a result, the input sentence is divided into the smallest meaningful units (morphemes). For example, each word is classified and extracted (divided) into parts of speech such as nouns, verbs, particles, auxiliary verbs, and adjectives. In this embodiment, morphological analysis is performed. However, there are various types of analysis, such as analysis for matching with a predetermined keyword or character string pattern, and analysis for comparing with a word registered in a dictionary. It is possible to analyze the input sentence by the method and extract a plurality of words.

At this time, words that are not directly related to the structure of the generated search sentence are deleted or adjusted. For example, in the case of an input sentence such as “What is a non-bitter medicine?”, “Ha”, “what”, “?”, Etc. are not reflected in the search sentence, and are deleted at this stage. In addition, nouns, adjectives, and the like are converted into a form (style) suitable for use in a search sentence as necessary.

Next, in step S33, the word corresponding to the operator included in the search sentence is determined from the analysis result (divided word) of the input sentence. At this time, the target word related to the word corresponding to this operator is determined. Here, the target word is a word indicating an element necessary for the calculation by the operator, and is a noun, an adjective or the like (basically) adjacent to the word determined to correspond to the operator. The target word can be defined as a search target item or search value of the search target data 400 in the search sentence.

For example, when continuous morphemes such as “other than” “no” are extracted from a part of the input sentence, the word “other” obtained by integrating these two words is a noun adjacent (for example, before and after). In consideration of a target word made up of, etc., it is determined to correspond to NOT of the logical operation. Furthermore, when a morpheme such as “de” is extracted from a part of the input sentence, the word “de” corresponds to the AND of the logical operation in consideration of the target word composed of adjacent nouns etc. Is determined.

Furthermore, according to the present embodiment, if necessary, the analysis data 340 is referred to, whether or not the adjective is a negative form, a word corresponding to the operator is determined, and as a result, It is possible to include NOT, and based on the type of the target word adjacent to the word (for example, the type of part of speech), it is possible to determine which logical operation operator the word corresponds to.

With this configuration of the present invention, a specific character string in an input sentence is not simply extracted and associated with an operator, but the true intention or nuance of the input sentence is correctly extracted and reflected in the search sentence. be able to. For example, in the conventional technology, the operator “AND” is simply associated with the character string “to” of the input sentence, or the negative meaning in the search word “not gorgeous” is not reflected in the search sentence. However, in the present invention, as described above, the word corresponding to the operator and the target word related to the word are determined from the words divided into a plurality by the input analysis. This is performed in consideration of related target words, and as a result, the true intention and nuance of the input sentence are correctly reflected in the search sentence.

Next, in step S34, the frame of the search sentence is determined from the analysis result of the input sentence in step S32 and the operator determined in step S33. The search sentence frame defines a skeleton of the search sentence, and includes a parenthesis arrangement (search order evaluation order), an operator, and a variable (a target word is arranged later) in the search sentence.

Next, in step S35, based on the analysis result of the input sentence in step S32, the corresponding target word is placed at the position of the variable included in the frame of the search sentence, and the variable is replaced with the target word. Generate a statement.

Next, in step S36, the generated search text is transmitted to the information providing server 200.

As described above, in the search sentence generation server 300 according to the first embodiment, the received input sentence is analyzed, the word corresponding to the operator is determined, the frame of the search sentence is determined, and then the search sentence of the search sentence is determined. Replace the variable part of the frame with the target word to complete the final search sentence. With such a search sentence generation method, it is possible to generate a search sentence that reflects the true intention or nuance of the search included in the natural sentence, based on the natural sentence input by the user.

Note that a more specific method related to search sentence generation in the search sentence generation server 300 according to the first embodiment will be described later in detail with reference to FIG.

FIG. 4 shows an example of a graphical user interface (GUI) of the conversation screen 120 displayed on the display of the user terminal 100. The conversation screen 120 shows a character display unit 121 and an input response display unit 122. The character display unit 121 displays, for example, a concierge character that changes facial expressions and movements in response to input text input by the user 10 and display of search results. The input response display unit 122 displays a natural sentence input sentence input by the user 10 to the user terminal 100.

Furthermore, in the input response display unit 122, as a concierge response, a message indicating that a search based on the input sentence is being performed and a search sentence generated based on the input sentence are shown. Thereafter, the search result corresponding to the input sentence transmitted from the information providing system 1 is displayed as a concierge response at the bottom.

Further, as shown in FIG. 4, as a search result from the information providing system 1, a link button 122a for displaying the number of search hits and detailed information of the search result is shown. For example, when the user 10 presses (or touches) the link button 122a, the display of the user terminal 100 transitions to another WEB page that displays detailed information of the search result. On this WEB page, for example, a list of items that satisfy the conditions of the search statement and link information to data representing details of each item are displayed.

FIG. 5 explains the search sentence generation process executed by the search sentence generation server 300 according to the first embodiment in more detail using an example input sentence 350.

As shown in FIG. 5A, the search statement generation server 300 receives an input sentence 350 of “non-blue clothes made by S company” from the user 10 and performs morphological analysis on the input sentence 350 (FIG. 5A). 3 step S32). In FIG. 5A, the result of the morphological analysis is shown as an analysis result 351.

By the morphological analysis, the input sentence 350 is decomposed into, for example, the words “blue”, “other than”, “no”, “clothes”, “de”, “S company”, and “made”. Here, for example, the words “blue”, “other than”, “clothes”, and “S company” are each determined as a noun, and the words “no” and “de” are each determined as a particle (case particle), The word “made” is determined to be a suffix. The word “Company S” is determined as a noun representing the company name by the proper noun dictionary included in the analysis data 340.

Next, the word corresponding to the operator and the target word are determined (see step S33 in FIG. 3). First, from the analysis result 351, “no” as a noun and “no” as a particle are extracted, and the two words are connected (for example, by referring to the analysis data 340) to be an operator. It is determined that it corresponds to “NOT”. If the natural sentence is “Q2 other than P2,” the search expression is configured as “Q2 NOT P2” (reference numeral 353 in FIG. 5A). In such a configuration, Q2 and P2 are determined as target words related to the word “other than”. In this example, Q2 is a noun, and P2 is a noun or adjective that modifies Q2.

Next, the particle “de” is extracted from the analysis result 351, and this word corresponds to “AND”. If the natural sentence is “Q1 in P1,” the search expression is configured as “P1 AND Q1” (reference numeral 352 in FIG. 5A). In such a configuration, Q1 and P1 are determined as target words related to the word “de”. In this example, Q1 is a noun, P1 is an adjective expression, and they are connected in parallel by “AND”.

In this way, the search expressions “Q2 NOT P2” and “P1 AND Q1” are extracted from the analysis result 351 by determining the operator and the target word. These two search expressions are integrated and searched. The sentence frame 354 is determined as “(Y NOT X) AND Z” (see step S34 in FIG. 3). Here, from the arrangement relationship and contents of the words (“other than” and “de”) corresponding to the operator and the target words adjacent to (for example, before and after) included in the analysis result 351, “P1 "And" Q2 "correspond to the same noun" clothes ", and" Q2 "is further modified with" P2 "and the operator" NOT ".

As a result, since “P1” is associated with “Q2 NOT P2”, the frame 354 of the search sentence is first “(Q2 NOT P2) AND Q1”. Here, the search expression “Q2 NOT P2” is included in parentheses in accordance with the fact that “P1” is associated with “Q2 NOT P2”. Some logical operations are prioritized. Note that just because a logical operation has a high priority, the logical operation is not always executed first. This is because a search engine or the like that interprets a search sentence may be expanded into another search expression having the same meaning.

After that, by replacing with Q2 = Y, P2 = X, Q1 = Z, “(Y NOT X) AND Z” as described above is determined (Note that replacement with X, Y, Z as described above is Just to make the explanation easier to understand).

Next, as shown in FIG. 5B, based on the analysis result 351, when X, Y, and Z are applied, X = “blue”, Y = “clothes”, Z = “manufactured by S company”. The search statement 355 generated automatically is “(clothing NOT blue) AND made by S company”.

The noun “blue” decomposed in the analysis result 351 may be converted into a corresponding adjective “blue” (for example, obtained using a synonym dictionary included in the analysis data 340). it can. Here, the adjective expression “manufactured by S company”, which is a combination of the noun “S company” and the suffix “manufactured” decomposed in the analysis result 351, is associated with the above Z, but the noun “S company”. May be associated with Z.

Based on the input sentence related to the search, “clothes other than blue made by S company” input by the user 10 through the above-described search sentence generation processing, a search sentence such as “(clothes NOT blue) AND made by S company”. 355 will be generated.

With such a configuration of the present invention, a specific character string in an input sentence is not simply associated with an operator, and the true intention or nuance of the input sentence is correctly extracted and reflected in the search sentence. Can do.

FIG. 6 summarizes various cases of the search statement generation processing described with reference to FIG. An operator and a target word are determined based on a word determined in the input sentence and a word adjacent to the word, and a search sentence is finally generated. However, the search statement generation processing of the present invention is not limited to the cases listed here.

Example of Case 1 relates to the operator of the logical operation “NOT”, and this operator represents “Negation”. When the example of the input sentence is “non-bitter medicine”, the word “bitter” of the target word is determined as an adjective (consecutive form) and the word “not” is determined as an adjective (auxiliary adjective) by morphological analysis. The word “medicine” is determined as a noun. Here, the word corresponding to the operator is “none”. In this case, the word “absent” is determined to have the meaning of negating the adjective “bitter”, and this word is determined to correspond to the operator “NOT”. After that, “drug” and the end of the adjective “bitter” “bitter” are selected as target words, and a search sentence “drug NOT bitter” is generated.

Example of Case 3 relates to an operator of logical operation “OR”, and this operator represents “logical sum”. When the example of the input sentence is “red or blue clothes”, the word “red” of the target word is determined as an adjective, the word “or” is determined as a conjunction, and the target word “blue” is analyzed by morphological analysis. Is determined as an adjective. Here, the word corresponding to the operator is “or”. In this case, the word “or” is determined to mean a logical sum as described above, and this word is determined to correspond to the operator “OR”.

Here, for example, partial search sentences “red OR”, “OR blue”, “blue AND clothes” are generated, and these partial search sentences are integrated to generate a final search sentence (here, In the input sentence “blue clothes”, “blue” and “clothes” are the target words, and the word corresponding to the operator is “blue” that modifies “clothes”). In addition, the partial search sentence “blue AND clothes” described above modifies the target word “clothes” with the operator “AND” and the target word “blue”. The partial search sentences “red OR” and “OR blue” ”Is integrated with the search sentence“ red OR blue ”in parentheses so that“ red OR blue ”is prioritized in the logical operation. Such use of parentheses in a search sentence is determined from the arrangement relationship and contents of a word corresponding to an operator and a target word adjacent to the word.

Cases

2, 4 to 6 also relate to operators of logical operations, and as in case 1, search sentences are assembled based on words corresponding to the operators included in the input sentence. Here, detailed description is omitted.

Example of Case 7 relates to the operator of the comparison operation “==”, and this operator represents “equal”. For example, when the input sentence is “price is 10,000 yen”, the word corresponding to the operator is “ga” as a particle (case particle). In this case, the input sentence is divided into “price”, “ga”, “1”, “ten thousand”, “yen” by morphological analysis, and “price” is adjacent to the word “ga” corresponding to the operator. ”And the number“ 10,000 ”indicating the quantity (here, the number and the number representing the unit are combined), the word“ ga ”becomes the operator“ == ”. It is determined that it corresponds. Thereafter, “price” is converted to “price”, “10,000” is converted to “10000”, each is selected as a target word, and a search sentence “price == 10000” is generated.

Examples of cases 7 to 12 also relate to operators for comparison operations, and as in case 6, search sentences are assembled based on words corresponding to the operators included in the input sentence. Here, detailed description is omitted.

Also, a search can be performed by combining a plurality of logical operations and comparison operations. For example, when the input sentence is “a non-bitter tablet medicine”, the search sentence can be generated as “drug AND tablet NOT bitter”.

Furthermore, a search can be performed by combining a plurality of logical operations and comparison operations. For example, if the input sentence is “red clothes whose price is 10,000 yen or less”, the search sentence can be generated as “clothes AND red AND price <= 10000”.

FIG. 7 is a flowchart showing the flow of a search statement generation process executed by the search statement generation server 300 'according to the second embodiment of the present invention. The search sentence generation server 300 ′ can be arranged in the information providing system 1 shown in FIG. 1 instead of the search sentence generation server 300 according to the first embodiment of the present invention.

Similar to the search statement generation process of the search statement generation server 300, the search statement generation process executed by the search statement generation server 300 ′ receives the input sentence of the user 10 from the information providing server 200. Based on the above, a search sentence is generated according to a predetermined rule, and the generated search sentence is transmitted to the information providing server 200.

First, in step S51, the search statement generation server 300 'determines whether or not the input statement of the user 10 has been received from the information providing server 200. When the input sentence has not been received (NO in step S51), the determination process is repeated. When the input sentence has been received (YES in step S51), the process proceeds to step S52.

In step S52, the search sentence generation server 300 'analyzes the received input sentence and extracts a plurality of words from the input sentence. In the present embodiment, morpheme analysis is performed using a morpheme database included in the analysis data 340. As a result, the input sentence is divided into the smallest meaningful units (morphemes). For example, each word is classified and extracted (divided) into parts of speech such as nouns, verbs, particles, auxiliary verbs, and adjectives. In this embodiment, morphological analysis is performed. However, the input sentence is analyzed by various methods such as analysis for matching with a predetermined keyword and analysis for comparing with a word registered in a dictionary. Can be extracted and a plurality of words can be extracted.

At this time, words that are not directly related to the structure of the generated search sentence are deleted or adjusted in the same manner as the search sentence generation server 300. In addition, nouns, adjectives, and the like are converted into a form (style) suitable for use in a search sentence as necessary.

Next, in step S53, it is determined whether or not an unprocessed word (word corresponding to the operator) exists in the analysis result (divided word) of the input sentence. If there is no unprocessed word (NO in step S53), the generation of the search expression has been completed. Therefore, in step S60, the generated search sentence is transmitted to the information providing server 200, and the process of step S51 (information The process of determining whether or not the input sentence of the user 10 has been received from the providing server 200 is repeated.

If there is an unprocessed word (YES in step S53), in step S54, the word corresponding to the operator included in the search sentence is sequentially determined from the analysis result (divided word) of the input sentence. For example, when continuous morphemes such as “other” and “no” are extracted from a part of the input sentence, the word “other” obtained by integrating these two words is the target word composed of adjacent nouns and the like. In consideration, it is determined that it corresponds to NOT of the logical operation. Furthermore, when a morpheme such as “de” is extracted from a part of the input sentence, the word “de” corresponds to the AND of the logical operation in consideration of the target word composed of adjacent nouns etc. Is determined.

Furthermore, according to the present embodiment, if necessary, the analysis data 340 is referred to, whether or not the adjective is a negative form, a word corresponding to the operator is determined, and as a result, Notation can be included, or based on the type of the target word adjacent to the word, it can be determined which logical operation operator the word corresponds to.

Next, in step S55, the target word corresponding to the operator is determined from the analysis result of the input sentence in step S52 and the one operator determined in step S54. Here, the target word is a noun, an adjective, or the like adjacent to the word determined to correspond to the operator, and corresponds to the search target item or the search value of the search target data 400 in the search sentence.

Next, in step S56, search sentences are generated and integrated based on the one operator determined in step S54 and the target word determined in step S55. Note that steps S54 to S56 are repeated while there are unprocessed words (words corresponding to operators) in the analysis result (divided words) of the input sentence. Each time a word corresponding to the word appears, it is sequentially integrated with the generated search sentence.

Next, in step S57, it is determined whether there is still a target word related to one operator determined in step S54. When there is no target word (NO in step S57), the process returns to step S53, and whether or not there is a further unprocessed word (word corresponding to the operator) in the input sentence analysis result (divided word). Determine whether.

If there is a target word (YES in step S57), in step S58, the target word corresponding to the operator is determined from the analysis result of the input sentence in step S52 and the one operator determined in step S54. Next, in step S59, the search sentences are integrated based on the one operator determined in step S54 and the target word determined in step S58. After step S58, the process returns to step S57, and it is determined whether there is still a target word related to one operator determined in step S54.

Thus, in the search sentence generation server 300 ′ according to the second embodiment, the received input sentence is analyzed, the words corresponding to the operator are sequentially determined, and each time the target word of the operator is determined, A search sentence is assembled based on the operator and the target word, and this process is repeated for all the words corresponding to the operator to complete the final search sentence. With such a search sentence generation method, it is possible to generate a search sentence that reflects the true intention or nuance of the search included in the natural sentence, based on the natural sentence input by the user.

FIG. 8 illustrates the search sentence generation process executed by the search sentence generation server 300 ′ according to the second embodiment in more detail using an example input sentence 360.

As illustrated in FIG. 8A, the search statement generation server 300 ′ receives an input sentence 360 “from clothes other than blue and made by S company” from the user 10, and executes morphological analysis on the input sentence 360 ( (See step S52 in FIG. 7). In FIG. 8A, the result of the morphological analysis is shown as an analysis result 361.

Through the morphological analysis, the input sentence 360 is decomposed into, for example, the words “blue”, “other than”, “no”, “clothes”, “de”, “S company”, and “made”. Here, for example, the words “blue”, “other than”, “clothes”, and “S company” are each determined as a noun, and the words “no” and “de” are each determined as a particle (case particle), The word “made” is determined to be a suffix. The word “Company S” is determined as a noun representing the company name by the proper noun dictionary included in the analysis data 340.

Here, the first word that is the operator of the search sentence is determined and acquired (see step S54 in FIG. 7). First, from the analysis result 361, the noun “other” and the particle “no” are extracted, and the two words are connected (for example, by referring to the analysis data 340) so that the operator It is determined that it corresponds to “NOT”. Next, a target word corresponding to the operator “NOT”, that is, a word adjacent to (for example, before and after) a word other than “not” is obtained from the analysis result 361 and determined as a target word (FIG. 7). Step S55).

As a result of this processing, the word “other than” corresponding to the operator “NOT” and the target word “blue” corresponding to the word “other than” are determined. The determination result 362 of FIG. A set of words other than "" is shown. Further, a search sentence “NOT blue” is generated based on the set of words other than “blue” (see step S56 in FIG. 7), and this search sentence is (1) of the partial search sentence 363 in FIG. 8A. It is expressed as

Thereafter, it is determined whether or not there is still a target word corresponding to the word “other than” (see step S57 in FIG. 7). As a result, the target word “clothes” is determined, and the determination result in FIG. 362 shows a set of words other than “clothes”. Further, a search sentence “clothes NOT” is generated based on the set of words other than “clothes” (see step S56 of FIG. 7), and this search sentence is the partial search sentence 363 of FIG. 8A (2). It is expressed as

Here, (1) and (2) of the partial search sentence 363 are further integrated as shown in FIG. 8B (see step S56 in FIG. 7), and “clothes NOT blue” is generated as the search sentence 364.

Thereafter, it is determined whether or not there is still a target word corresponding to the word “other than” (see step S57 in FIG. 7). As a result, since the target word does not exist, it becomes an operator of the search sentence. The next word is determined and acquired (see step S54 in FIG. 7). Here, the particle “de” is extracted from the analysis result 361, and the word “de” corresponds to the operator “AND”. Next, an object word corresponding to the operator “AND”, that is, a word adjacent to the word “de” is obtained from the analysis result 361 and determined as an object word (see step S55 in FIG. 7).

As a result of this processing, the word “de” corresponding to the operator “AND” and the target word “clothes” corresponding to the word “de” are determined, and the determination result 362 of FIG. A set of words is represented. Further, a search sentence “clothes AND” is generated based on the set of words “clothes” (see step S56 in FIG. 7). It is represented.

Further, (1), (2), and (3) of the partial search sentence 363 are integrated as shown in FIG. 8B (see step S56 in FIG. 7), and “(clothing NOT blue) AND is used as the search sentence 365. Is generated. Such integration processing is performed by integrating the search sentence 364 shown in FIG. 8B and (3) of the partial search sentence 363 shown in FIG. 8A and grouping them together with the common target word “clothes”. In addition, “clothes NOT blue” in the search sentence 364 modifies the subject word “clothes” with the operator “NOT” and the target word “blue”, and the search sentence “clothes” that is a partial search sentence. When integrated with “AND”, as described above, “clothes NOT blue” is included in parentheses so that it is prioritized in the logical operation. Such use of parentheses in a search sentence is determined from the arrangement relationship and contents of a word corresponding to an operator and a target word adjacent to the word.

Thereafter, it is determined whether or not the target word corresponding to the word “de” still exists (see step S57 in FIG. 7). As a result, the target words “S company” and “manufactured” are determined. In the determination result 362 of FIG. 8A, a set of words “made by S company” is represented. In the present embodiment, “Company S” and “Made” are extracted as morphemes in the analysis result 361, but here “S Company” and “Made” are used to make the target word an adjective expression. Are combined to make “S”. Of course, only “S company” which is a noun can be selected as the target word.

Further, based on the set of words “manufactured by S company”, a search sentence “AND made by S company” is generated (see step S56 in FIG. 7), and this search sentence is converted to the partial search sentence 363 in FIG. 8A. It is expressed as (4).

Further, (1) to (4) of the partial search sentences 363 are integrated as shown in FIG. 8B (see step S56 in FIG. 7), and “(clothing NOT blue) AND S company” is used as the search sentence 366. Is generated. Such integration processing is performed by integrating the search statement 365 shown in FIG. 8B and (4) of the partial search statement 363 shown in FIG. 8A and grouping them together in a common operator “AND” portion.

Thereafter, it is determined whether or not the target word corresponding to the word “de” still exists (see step S57 in FIG. 7). As a result, since the target word does not exist, the next as an operator of the search sentence. The word is determined and acquired (see step S54 in FIG. 7). Here, since there is no unprocessed word corresponding to the operator from the analysis result 361, the search statement 366 "(clothing NOT blue) AND S company" generated by the integration process so far is generated. It is transmitted to the information providing server 200 as a search sentence (see step S60 in FIG. 7).

In the search statement generation process executed by the search statement generation server 300 ′ according to the second embodiment of the present invention, as shown in FIGS. 7 and 8, the entire input sentence is analyzed at once (see FIG. 7 (step S52), and then the search sentence generation processing is performed in order for each word corresponding to the operator (step S53 to step S59 in FIG. 7), while partially analyzing the input sentence, When a word corresponding to the operator is extracted, it is possible to perform control so that a search sentence generation process related to the word is performed each time.

Next, with reference to FIG. 9, an example of a hardware configuration of a computer constituting the user terminal 100 according to the embodiment of the present invention will be described. The user terminal 100 is a portable device such as a smartphone, for example. However, the configuration of the user terminal 100 illustrated in FIG. 9 is merely an example of a typical configuration.

The user terminal 100 includes a CPU (Central Processing Unit) 101, a memory 102, a camera 103, a GPS control unit 104, a GPS antenna 105, a radio signal processing unit 106, a radio communication antenna 107, an audio control unit 108, a microphone 109, and a speaker. 110, a display control unit 111, an input device interface 112, a touch screen 113, a non-contact IC card reader / writer 114, a sensor 115, an auxiliary storage device 116, and an external recording medium interface 117.

The CPU 101 controls the operation of each component of the user terminal 100 and executes each function under the control of the OS.

The memory 102 is usually composed of RAM (Random Access Memory). In the memory 102, a program for realizing each function executed by the CPU 101 is loaded at the time of execution, and data necessary for the program is temporarily stored.

The camera 103 is a small imaging device built in the user terminal 100. The GPS control unit 104 receives signals from a plurality of GPSs via the GPS antenna 105 and checks the position of the user terminal 100.

The wireless signal processing unit 106 communicates with the mobile phone base station via the wireless communication antenna 107, transmits / receives call data to / from other devices, and communicates with other terminals (via the Internet). Controls transmission / reception of WEB pages and mail data. The wireless signal processing unit 106 implements wireless LAN communication with the wireless LAN access point using the wireless communication antenna 107, and performs data transmission / reception via the Internet (here, for convenience, the wireless signal processing unit 106 and the antenna 107 for wireless communication perform wireless communication with the mobile phone base station and wireless communication with the wireless LAN access point).

The audio control unit 108 controls the microphone 109 and the speaker 110 to realize a telephone call by wireless communication. On the other hand, the audio control unit 108 controls to output sound when playing a video or music in the application.

The touch screen 113 displays information on a display device configured by, for example, an LCD (Liquid Crystal Display), and the position where the user touches (presses) the screen surface with a finger or the like is displayed on the touch screen 113. Detect with capacitive type touch sensor. The display control unit 111 processes drawing data issued by the CPU 101 and outputs, for example, a WEB page, a moving image, or the like to the display device of the touch screen 113. In the input device interface 112, the touch sensor of the touch screen 113 detects an operation by the user, and transmits this to the CPU 101 as a predetermined signal.

The non-contact IC card reader / writer 114 stores data stored in the IC chip based on a command from the CPU 101 when the card in which the non-contact IC chip is embedded is held over a predetermined position of the user terminal 100. Or write predetermined data to the IC chip.

The sensor 115 is a sensor other than the touch sensor of the touch screen 113, and includes a motion sensor, a light sensor, a proximity sensor, and the like.

The auxiliary storage device 116 is composed of, for example, a semiconductor memory called a flash memory or a hard disk. The auxiliary storage device 116 stores various data as well as programs for realizing the functions executed by the CPU 101.

The external recording medium interface 117 accesses the external recording medium 140 and reads data recorded therein. The external recording medium 140 is, for example, a portable flash memory.

Next, with reference to FIG. 10, an example of a hardware configuration of a computer constituting the search statement generation server 300 according to the first embodiment of the present invention will be described. However, the configuration of the search statement generation server 300 illustrated in FIG. 10 is merely an example of a typical configuration. Such a computer configuration may be employed in the search statement generation server 300 ′ according to the second embodiment of the present invention, the information provision server 200 of the information provision system 1, and the like.

The search statement generation server 300 includes a CPU 301, a RAM 302, a ROM (Read Only Memory) 303, a network interface 304, an audio control unit 305, a microphone 306, a speaker 307, a display controller 308, a display 309, an input device interface 310, a keyboard 311 and a mouse. 312, an external storage device 313, an external recording medium interface 314, and a bus 315 that connects these components to each other.

The CPU 301 controls the operation of each component of the search statement generation server 300, and controls the execution of each function according to the present invention under the control of the OS.

The RAM 302 temporarily stores programs for executing each process executed by the CPU 301 and data used during the execution of these programs. The ROM 303 stores a program executed when the search statement generation server 300 is activated.

The network interface 304 is an interface for connecting to the network 320. The network 320 is a network for connecting to the information providing server 200, for example.

The audio control unit 305 controls the microphone 306 and the speaker 307 to control audio input / output. The display controller 308 is a dedicated controller for actually processing a drawing command issued by the CPU 301. The display 309 is a display device configured by an LCD, for example.

The input device interface 310 receives a signal input from the keyboard 311 or the mouse 312 and transmits a predetermined command to the CPU 301 according to the signal pattern.

The external storage device 313 is, for example, a storage device such as a hard disk or a semiconductor memory, and the above-described program and data are recorded in this device.

The external recording medium interface 314 accesses the external recording medium 330 and reads data recorded therein. The external recording medium 330 is, for example, a portable flash memory. A program executed by the CPU 301 to realize each function of the present invention can be provided from the external recording medium 330 via the external recording medium interface 314. As another distribution form of the program for realizing each function of the present invention, a route that is stored in the external storage device 313 or the RAM 302 via a network 320 and a network interface 304 from a predetermined server on the network. Etc. are also conceivable.

Until now, the search statement generation system 11 including the search statement generation server 300 according to the first embodiment of the present invention or the search statement generation server 300 ′ according to the second embodiment of the present invention, and the search statement generation system. 11 has been described with reference to specific examples in which the present invention can be implemented. However, these specific examples are merely examples for explaining the present invention, and the scope of rights of the present invention is limited. It is not limited to these specific examples. The technical idea of the present invention can be realized by various methods and configurations other than these.

DESCRIPTION OF SYMBOLS 1 Information provision system 11 Search sentence generation system 100 User terminal 200 Information provision server 300 Search sentence generation server 300 'Search sentence generation server 340 Analysis data 400 Search object data 500 Network

Claims

An input sentence receiving means for receiving an input sentence of a natural sentence input from a user;
An input sentence analyzing means for analyzing the input sentence and dividing it into a plurality of words;
Determining means for determining a word corresponding to an operator based on the plurality of words and determining a target word related to the determined word;
Search sentence generation means for generating a search sentence corresponding to the input sentence by using an operator corresponding to the determined word and a set of target words determined to be related to the word. Search sentence generation system.
2. The search sentence generation system according to claim 1, wherein the determination means determines a word corresponding to an operator of a negative logical operation based on a predetermined determination criterion.
The search sentence generation means uses a predetermined operator based on a relationship between the operator corresponding to the determined word and the operator and the target word in each set of target words determined to be related to the word. The search statement generation system according to claim 1, wherein the search statement is generated so that a priority of a logical operation is determined.
The determination means sequentially determines words corresponding to operators for the plurality of words,
The search sentence generation unit generates a search sentence relating to the determined word each time a word corresponding to the operator is determined, and integrates the search sentence with the generated search sentence. The search sentence generation system according to claim 1.
A search statement generation method that is executed in a search statement generation server that generates a search statement based on an input sentence of a natural sentence input by a user,
An input sentence receiving step of receiving the input sentence;
An input sentence analysis step of analyzing the input sentence and dividing it into a plurality of words;
A determination step of determining a word corresponding to an operator based on the plurality of words and determining a target word related to the determined word;
A search sentence generation step of generating a search sentence corresponding to the input sentence by using an operator corresponding to the determined word and a set of target words determined to be related to the word. Search sentence generation method.